Research And Development Of Key Technologies On Key Words Extraction And Sentiment Classification Of Video Website Comment

In recent years,with the huge growth of online video,more and more users began to post comments on the video site.Users' comments usually include personal emotions and some of the key information about the video,imposing a significant impact on video viewing decisions for the other users.Classification and extracting keywords from online video comments automatically have become an urgent problem.This thesis focuses on the emotional classification and keyword extraction of online video comments.In the aspect of emotion classification,this thesis expounds the specific methods of crawling online video comments and puts forward the method of auto-tagging corpus based on emotion dictionary.In the process of constructing the emotion classifier,a machine learning model was constructed.This thesis extracted the words,the collocation of the two words and the combination of the two as the feature respectively,and used the mutual information and the chi-square statistics as the feature selection method.In the selection of classification algorithm,Naive Bayesian,Logistic Regression and Support Vector Machine were used.The influence of different feature extraction and selection methods and different classified algorithms on the accuracy of emotion classification was analyzed.Based on the above steps,the emotional classification model of online videos' comments is finally determined.In terms of keyword extraction,this thesis expatiated on the principle and usage of the Chinese keyword extraction algorithm.TextRank algorithm has been developed and used to extract keywords from the network video comments.This thesis not only extracts keywords,but also concretely judges the emotional polarity of the keywords,and counts the keywords according to the different time granularity.Finally,the emotion classification model and the keyword extraction have been tested in the specific scene.